SiFive has raised $400 million in an oversubscribed Series G round, lifting its valuation to $3.65 billion and giving the company more firepower for its push into high-performance RISC-V data center CPUs and AI infrastructure. Arm and Intel are both making the case that CPUs remain central to AI systems, even in an accelerator-led market. For SiFive, the raise looks less like bridge financing and more like a declaration that it intends to become the leading independent RISC-V IP supplier, with an eventual IPO more plausible than a trade sale.
(Source: SiFive)
SiFive has secured $400 million in new funding, in a round led by Atreides Management with participation from Apollo Global Management, Nvidia, Point72 Turion, T. Rowe Price Investment Management, Prosperity7 Ventures, and Sutter Hill Ventures—an impressive crowd to say the least. The company said the oversubscribed Series G round values SiFive at $3.65 billion and will be used to accelerate development of high-performance RISC-V data center solutions, including scalar, vector, and matrix compute IP, software enablement, and customer programs around AI infrastructure.
The announcement lands at a moment when the CPU has regained strategic weight in AI infrastructure. Arm recently moved beyond IP licensing with its Arm AGI CPU, positioning that silicon as a foundation for agentic AI systems and drawing public support from both Meta and OpenAI around the orchestration layer of large-scale AI workloads. Intel, on the same day as SiFive’s funding news, expanded its collaboration with Google, saying Xeon processors will continue powering Google Cloud infrastructure across AI, inference, and general-purpose workloads while the two companies deepen co-development of custom ASIC-based IPUs. The message from both is similar: Accelerators may dominate attention, but CPUs are important (again).
SiFive’s pitch looks better timed than it might have a year ago. The company is no longer talking like a RISC-V vendor hoping to ride general ecosystem enthusiasm. It is talking like a company that wants to supply a serious slice of AI data center compute and, more specifically, the CPU layer that coordinates increasingly heterogeneous systems. Its earlier NVLink Fusion deal with Nvidia already pointed in that direction, giving SiFive’s RISC-V compute platforms a route into coherent, high-bandwidth AI fabrics rather than leaving the architecture confined to embedded and edge narratives.
EE Times reported SiFive as saying it is “working with several hyperscaler customers on future high-performance CPU design requirements,” and that all of them are already existing SiFive customers in other performance segments. That still leaves SiFive without the kind of named hyperscaler attachment that Arm now enjoys with Meta, or Intel with Google, but it suggests traction behind the scenes. We are also aware of projects at some very large players looking at RISC-V for AI data center CPU.
This is also consistent with what John Ronco told us in our January interview. He said SiFive aimed to be “the gold standard, offering a complete portfolio from low-end microcontrollers to high-performance cores for servers.” He also stressed that the company still prioritizes configurability, even as some rivals move toward more fixed subsystems. In other words, SiFive’s strategy has been to remain broad in portfolio while pushing harder into premium compute.
We don’t believe SiFive plans to significantly shift from IP towards silicon, as Arm has done and Semidynamics is doing. Ronco told us that silicon would remain the partner’s role.
What do we think?
The most interesting thing about this investment is not the size, though $400 million is substantial. It is what the raise says about where SiFive thinks the market is going. Arm’s latest move and Intel’s latest Google tie-up both reinforce the same idea: The AI data center is not a GPU-only story. CPUs remain the anchor for broader heterogeneous systems. SiFive wants to claim part of that layer before the market structure hardens around a few incumbents.
This funding also pushes SiFive further outside the current M&A wave in RISC-V. We have already seen consolidation pressure build across the sector, and our own prior reporting has argued that there are simply too many IP vendors chasing too little durable value. Ronco himself told us that consolidation is inevitable and that, in the end, the market will support only a few high-quality IP vendors alongside in-house efforts and niche players. A company valued at $3.65 billion is not impossible to buy, but in the current climate, it is hard to see how an acquirer could pay enough to satisfy investors without stretching logic. That makes SiFive look less like a trade-sale candidate and more like a company trying to grow into a public-markets story.
Longer term, we think IPO is the path for the company. That fits the valuation, the continued fundraising, and the company’s positioning as a category leader. SiFive appears to be trying to maximize revenue growth and strategic relevance, not optimize itself for a near-term sale. China will likely remain part of that equation, because openness, configurability, and the search for alternatives to incumbent architectures continue to matter there, but the more immediate narrative is now AI data center CPU.
There is a missing component in SiFive’s offer: GPU. Ronco was direct with us on that point: “GPUs are a whole conversation in themselves. While it’s possible to build a GPU with RISC-V processors and accelerators, it’s not a priority for us. We’re content to let others solve the GPU problem. Instead, we’re focused on AI and Tensor Cores.” That is perfectly reasonable in the short term, especially when Nvidia is providing the accelerator and fabric side of the story. But strategically, it leaves a hole. If SiFive wants to become the largest and most complete RISC-V infrastructure IP company, eventually the absence of a GPU story will start to matter more, not less.
Historically, SiFive has worked with Imagination on that missing piece. Imagination joined SiFive’s DesignShare ecosystem in 2019 to make PowerVR GPU and NNA IP available to SiFive users, and more recently, Imagination GPU IP has appeared alongside SiFive CPU IP in RISC-V edge SoCs such as Eswin’s EIC77 family. That does not mean anything larger is underway, but it does make a deeper tie-up imaginable. A merger between SiFive and Imagination would be speculative, and we are not suggesting one is imminent. But at a strategic level, it would make more sense than many other combinations in the current RISC-V market. SiFive brings the leading CPU franchise and a growing AI data center narrative. Imagination brings mature GPU IP, software, and long experience selling graphics into embedded systems. In a market moving toward more complete heterogeneous platforms, that kind of complementarity is more valuable than simply acquiring another CPU house with overlapping ambitions.
However, SiFive is not without SIMD or SIMT capabilities, having recently adopted RISC-V Vector extensions with BF16 and FP8.
For now, though, SiFive’s immediate task is simpler and harder at the same time: turn capital, ecosystem credibility, and whatever hyperscaler engagement it has, into revenue growth. The company has bought itself time and room to do that. More importantly, it has aligned its story with a wider market realization that the CPU is back in fashion in AI infrastructure. If that trend holds, SiFive may yet become not just the best-known RISC-V IP company, but the largest one.
LIKE WHAT YOU’RE READING? TELL YOUR FRIENDS; WE DO THIS EVERY DAY, ALL DAY.